DISCO Nets: DISsimilarity COefficient Networks
This addresses the need for better uncertainty estimation in deep learning, offering a novel approach for tasks requiring probabilistic predictions.
The paper tackles the problem of probabilistic modeling with neural networks by introducing DISCO Nets, which efficiently sample from a posterior distribution and are trained by minimizing a dissimilarity coefficient, resulting in outperforming non-probabilistic networks and existing probabilistic models in modeling uncertainty.
We present a new type of probabilistic model which we call DISsimilarity COefficient Networks (DISCO Nets). DISCO Nets allow us to efficiently sample from a posterior distribution parametrised by a neural network. During training, DISCO Nets are learned by minimising the dissimilarity coefficient between the true distribution and the estimated distribution. This allows us to tailor the training to the loss related to the task at hand. We empirically show that (i) by modeling uncertainty on the output value, DISCO Nets outperform equivalent non-probabilistic predictive networks and (ii) DISCO Nets accurately model the uncertainty of the output, outperforming existing probabilistic models based on deep neural networks.